Amps and ohms part for DIY DMM

[Kleinstein]

I got a DVM part of a DIY DMM roughly working (https://www.eevblog.com/forum/metrology/diy-high-resolution-multi-slope-converter/msg3827432/#msg3827432). The SW and protection still needs some work. Here is now my idea for the amps and ohms part.

The main part is rather conventional: amps range switching is with latching relays and the classic protection with fuse and diode bridge.
The low amps part is using a trans-impedance amplifier, with a slightly unusual 2 OP configuration to use a 5 V powered AZ OP. 

The ohms current source is also relatively conventional:
The positive current source is based on 4 resistors and 3 voltages ( -5 or -1.5 or -0.5 V relativ to a 12 V upper end to choose on the main part) and a P MOSFET for control.
The protection is similar to the HP meters with cascaded PNPs. Switching to the terminals for the ohms part is with reed relays (in hope for low leakage).
For this main part the question in mainly if there is something wrong by a stupid mistake. For the Ohms source I am not sure if the extra diodes in series to the switches (e.g. like in the 3458 or Keithley 2002) are worth it. One can get away without it for the largest resistor, as this path is active when the low currents are used. For the highest currents the transistor acts as diode, when off.

There are 3 more unusual parts:
1) The amps mode uses 2 extra terminals and these are fully separated, when in voltage mode. This is done to allow electronic switching also on the low side for the voltage measurement and the ohms mode. So there is no simple common terminal, but a separate amps low side. The direct link, bypassing all the shunts is needed anyway to allow for an internal measurement of the shunts, even with parts connected.

2) There is some extra circuit to limit the open circuit voltage for the ohms source. I think this makes sense, as the open circuit voltage could be quite high (e.g. up to some 20 V), possibly too high for some DUTs. The buffer amplifier can also be used for guarding some of the parts and to use the ohms source as an extra voltage input, with very low bias, though some drift.
The new part is the voltage limiting part around U9B. This part is a bit uncertain with the reaction speed and leakage.

3) The ohms part can also be switched as a trans-impedance amplifier to measure small current (e.g. up to some 2 µA). This is a low effort addition with the amplifier part already there.

[2N3055]

Voltage limiting (programing ?) circuit is a good idea.
I would just like to add that on Metrix I have high voltage/current mode (21V/10mA) that is very useful for small voltage zeners and multiple serial connected LEDS /multichip module LEDS.
Having control of both current and compliance voltage would be very versatile for both resistance mode and diode mode...

[Kleinstein]

The voltage limit is essentially only to protect the DUT. 10 mA at 20 V may damage some small resistors or LEDs in revers. 20 V is nice for high level resistors and in my case the relatively high voltage carries over from the main voltage range.
Otherwise a fine adjustable voltage limit in ohms more is of limited values. It may work in current mode to do some primitive curve tracing. So set a voltage and than read the current. Maybe it could be worth getting a slighly lower noise DAC ? (e.g. the version with external ref.)

[2N3055]

The voltage limit is essentially only to protect the DUT. 10 mA at 20 V may damage some small resistors or LEDs in revers. 20 V is nice for high level resistors and in my case the relatively high voltage carries over from the main voltage range.
Otherwise a fine adjustable voltage limit in ohms more is of limited values. It may work in current mode to do some primitive curve tracing. So set a voltage and than read the current. Maybe it could be worth getting a slighly lower noise DAC ? (e.g. the version with external ref.)

Absolutely agree. 20V/10mA is a special mode you have to enable on Metrix but it is great to have if you need it. Otherwise it is dangerous. It can kill quite a few low voltage components. Metrix shows a warning when you enable it.

Fine control I also don't think is very useful. For diode mode few discrete values would be enough.

Low voltage/ higher current mode for continuity would be interesting. Like 200-300 mV. So semiconductors wouldn't turn on.
Simple curve tracer is an interesting idea though.

But that is getting into specialized instrument area.... Maybe a bit too much of a tangent from a just multimeter.
Or not. I'm sharing my thoughts with you about what I would improve in the meters I have and tried.
They usually have too little control of and limited continuity test. Limited diode test with not enough control over voltage/current. Generally not fully optimized low ohm tests.

[Kleinstein]

I think the curve tracing idea could be a software only addition in my case - it came to my mind only after the circuit was drawn. It is also limited to some 5-10 mV steps up to 10 V and around 5 mA max.
Otherwise I also don't see a need for fine steps on the ohms open circuit voltage. The main steps would be a very low one, like 50 mV for dry ohms,  maybe some 300 mV to not turn on a diode junction, some 4 V to be safe on LEDs and keep the power in limits and in reverse and than the full ~20 V.
20 V and 10 mA would be the exception. At 1 µA a maximum of 20 V makes sense for higher resistors.

For the limits the 2 options I considered are a cheap 10 bit DAC + gain and a MUX (e.g. DG408 or 4051) with a divider chain. The DAC + gain stage seem to be a bit smaller and more flexible.

[bsw_m]

Switching to the terminals for the ohms part is with reed relays (in hope for low leakage).

I recommend that you consider using the Panasonic AGN210 and Omron IM (IM41) relays - these relays have low leakage (lower than in the reed relays I tested), and reliable contact at low currents.

[Kleinstein]

Switching to the terminals for the ohms part is with reed relays (in hope for low leakage).

I recommend that you consider using the Panasonic AGN210 and Omron IM (IM41) relays - these relays have low leakage (lower than in the reed relays I tested), and reliable contact at low currents.

That is interesting.  I did not know those 2 relais types - they are really tiny and may be an alternative also for the amps part. The relays there ( Fujitsu FTR-C1 or Kemet EC2 are similar, but still a bit larger. 2 contacts in series would also give a higher votlage capability.
So the main advantage left for reed relays seems to be speed and low acoustic noise.

I already have the reed relais at hand and will not chance in this point. I don't expect performance deep in the pA range - if it works there it is nice, but no priority.

[bsw_m]

Some measurement data for AGN210 might be helpful.
Switching current 1pA:
https://pasteboard.co/jdsKQfo3wn9J.png

Leakage current at test voltage 50V, is 1E-16A.
Connection for leakage test:
https://pasteboard.co/nLaQy42ngR9r.jpg

[2N3055]

I think the curve tracing idea could be a software only addition in my case - it came to my mind only after the circuit was drawn. It is also limited to some 5-10 mV steps up to 10 V and around 5 mA max.
Otherwise I also don't see a need for fine steps on the ohms open circuit voltage. The main steps would be a very low one, like 50 mV for dry ohms,  maybe some 300 mV to not turn on a diode junction, some 4 V to be safe on LEDs and keep the power in limits and in reverse and than the full ~20 V.
20 V and 10 mA would be the exception. At 1 µA a maximum of 20 V makes sense for higher resistors.

For the limits the 2 options I considered are a cheap 10 bit DAC + gain and a MUX (e.g. DG408 or 4051) with a divider chain. The DAC + gain stage seem to be a bit smaller and more flexible.

That sounds absolutely perfect. And would be light years better than from any store bought meters out there..
I might be building one only because of that feature...

[Kleinstein]

I got a few first results from real hardware: The PCB is partially populated (no voltage limit and no amps part so far).
A few more minor changes to the circuit: C1 changed from 10 nF to 20 nF to make the bootstrapped buffer stable.
A few resistor values changed a little (e.g. 47 K instead of 22 K at the current sources) and a few more decoupling caps.
R32 should change to a larger value (have to still find a good one) around 680 ohms to avoid the bootstrapped buffer to get stuck at the positive rail.
The input amplifier for the BS buffer so far is a MCP6001 instead of the planed OPA377 (got the wrong package version ).

The first test is with the current source in the lower current ranges (1.5 M resistor active).
The current source at nominally 1 µA is directly send to the voltmeter part with the 10.5 M divider engaged and the voltage during warm up is measured. The curve uses some 1.5 readings per second (32 PLC). The current is reasonable stable, though some drift is visible and the noise is a bit higher than I had hoped for. To my surprise even the nearly off state (8 nA from some 12 mV accross the 1.5 M) case seems to be resonable stable and could be used to get a rough reading also for large resistors (up to ~ 2 Gohms).

[Kleinstein]

Added a few more parts and some more software. The Ohms current source seems to work now from 0.3 µA to 10 mA. The R+L filter at the output is kind of needed for stability and in my cases needed quite a large inductor. I am still not sure the current source is stable with a transformer DUT under all conditions. At least it works with the inductors / ferrite transformers I tested. The initial version even oscillated with a short to ground at some currents.
The ohms source stability against oscillation is definitely a point to whatch out for and not easy

The low current transimpedance amplifier part with MCP6001 (not an ideal choice, but the part I had at hand in the right form factor) and 5 M resistance works. As a test I checked the open contact leakage of a reed relay (Little fuse HE751A) with some +-8 V as the input voltage with the tia measurding the current. There is no offset subtracted, so most of the offset is from the OPs offset votlage, not actual current flow.
The curve shows some transients (likely dielectric absorbtion of the PCB and relay) and a difference in current of some 0.5 to 1 pA.
So the isolation of the reed relay is reasonable, but not perfect.

I still have a problem with the measurment of the higher currents of the ohms source: The low side has limited drive strength and can not provide the return current for the 3 mA and 10 mA test currents.  The problem is with the DVM board and not the new daughterboard.

[Kleinstein]

I finally got the buffer amplifier with bootstrapped supply working. To avoid the buffer to get stuck at the upper rail it took some modifications. The problem was that one of the diodes (D12) got reverse biased when at the positive rail and than the amplifier get stuck there. I can be resolved by limiting the maximum voltage the lower end of the bootstrapped supply can reach with 1 pair of zener diodes (parallel to C14) and than a resistor in parallel to the 2.7 V zener diode, so that even at the upper rail (with the current source at the edge) there is enough current flow for the load. The attached graphics shows measured voltages before the final added 10 K parallel to D10.

The buffer is now really high impedance / low bias. With a 100 pF cap at the input the drift rate is at a few mV/s and thus an input bias in the sub pA range. Leakage through the relays may be a large part of this. With only 1 reed relay (towards the shunts) installed the estimated impedance is in the 10-100 TOhms range.

So I finally got a first test on the ohms ranges in 4 wire mode. The DUT is a DIY wire wound resistor from manganin wire planed to use as one of the shunts.
There is still some drift. With the relatively small test current the noise ist still quite visible. Chances are part of the drift is from thermal EMF if the clamps used. The current source may also contribute to the drift.

I think I need a few more better resistors - especially the 50 K I tested was a total disaster: supposed to be 25 ppm/ K grade, but measured more like 100 ppm/K.
The low ohms bare wire shunt is also not working well. The TC may be OK, but quite some thermal EMF when heating 1 side more than the other.

For comparison I tested the manganin wire I used for the wire wound resistor for thermal EMF and this came out as < 0.5 µV/K against the copper wire I used for the test. So chances are the 10.7 Ohms are OK.  The TC also looks good, at least in a limited temperature range. A weak point with manganin the rather high 2nd order TC.

[Kleinstein]

There is a slight problem with the DVM part of the project that took some time:

The 200 mV range (gain of some 100) did not behave well with a high resistance (e.g. > 20 K) source and tends to oscillates. This seems to be a principle limitation on how the input stage is made with an moving low side and related to capacity at the input toward ground. Another part of the problem could be from OPs (currently TL032) that are a bit on the slow side and cause delay for the guard signal and driven supply. The faster amplifiers I had in mind (e.g. TL072H, OPA2991) are currently hard to get. So a fix of this part may take some time. It is currently not high priority for me.
Some capacitance could be removed, but not all. Also a few tweaks with the compensation of the amplifier helped a little.
I got it now working reasonable up to some 50 K - the higher resistance cases would need some 1-2 nF at the input, which is not really nice, but for the start acceptable.
Really low noise is important for low impedance sources. 100 K ohms already have more noise than the amplifier. Noise wise I actually don't need that much gain - already a gain of 10 is where the amplifier noise becomes more important than the ADC noise. So a slightly reduced gain can be an option too.

The oscillation problem surfaced when checking for current noise from the amplifier. In the lower gain mode it came out at some 85 fA/sqrt(Hz). That is quite a bit higher than the number in the DS for the AD8628, but still realistic. About half the current noise of the LTC2057 with about twice the voltage noise makes sense.
So current noise gets more important than voltage noise from some 300 K on for the AD8628. With the ohms mode the current source would give larger current noise from the amplifier (MCP6V66 with about 100 fA/sqrt(Hz)) and in parts from the resistors.

I finally got some more stable (and low tolerance) resistors to test the stability of the ohms source and for the low current measurement.
The residual drift seen is likely more from the current source than the resistor under test. The curve shown is more like a best case for the test, with a rather high test current (300 µA) and relatively high power (2.7 mW) for the DUT. Normally one would likely select a smaller test current (e.g. 100 µA or 30 µA). So far the noise looks good - though I am slightly missing a comparison.